Associations between the American College of Rheumatology pediatric response measures and the continuous measures of disease activity used in adult rheumatoid arthritis: A secondary analysis of clinical trial data from children with Polyarticular-Course Juvenile Idiopathic Arthritis

Authors


Abstract

Objective

To measure associations between the American College of Rheumatology (ACR) pediatric criteria for improvement and the continuous measures of disease activity used for rheumatoid arthritis in adult patients with polyarticular-course juvenile idiopathic arthritis (JIA).

Methods

In this retrospective analysis of 2 etanercept trials, disease activity was calculated at baseline, 3 months, and 6 months using the Disease Activity Score (DAS), the DAS based on 28 joints (DAS28), the Simplified Disease Activity Index (SDAI), and the Clinical Disease Activity Index (CDAI). The ACR pediatric response and the European League Against Rheumatism (EULAR) response were also determined for the 3-month and 6-month evaluations. Data were analyzed in 94 patients with JIA independent of the treatment arm. Correlation coefficients between measures were calculated for each visit. The areas under the receiver operating characteristic curve (AUC of ROC) were calculated to assess the discriminative properties of the scores for the ACR pediatric response measures.

Results

The mean DAS, DAS28, CDAI score, and SDAI score were 3.7, 4.7, 30.8, and 36.4, respectively, at baseline, corresponding to high levels of disease activity (CDAI/SDAI) or moderate levels of disease activity (DAS/DAS28). At 3 months, the mean scores corresponded to low (DAS/DAS28) or moderate (CDAI/SDAI) disease activity. At 6 months, the mean scores corresponded to low disease activity (DAS/DAS28/CDAI) or moderate disease activity (SDAI). Most children met the criteria for a good or moderate EULAR response at 3 months and 6 months. The correlation between continuous outcome measures and each pediatric core set component was moderate to very good. The AUC of ROC values for each measure were high (range 0.76–0.98).

Conclusion

Good correlation and discriminative abilities were seen between the DAS, DAS28, CDAI, and SDAI for the ACR pediatric criteria for improvement. These disease activity measures may be useful for research and clinical care in polyarticular-course JIA.

Juvenile idiopathic arthritis (JIA) is the most common pediatric rheumatic disease, with an estimated annual incidence of ∼3.2–6.1 cases per 100,000 persons, depending on the case definition used and the population studied (1–3). Polyarticular JIA, which accounts for ∼40% of cases of JIA, is particularly refractory to standard medical therapies. Previous observational studies have demonstrated that children with polyarticular JIA have only a 15% probability of achieving disease remission within 10 years of the initial diagnosis, and ∼65–70% of children with polyarticular-onset JIA enter adulthood with ongoing active disease (4–6). In addition, a recent retrospective cohort study showed that in the majority of children with polyarticular JIA, ≥60% of their disease course involved active disease (7).

However, outcomes in both observational and interventional studies of polyarticular JIA remain incompletely described, primarily due to limited data and to a lack of standardized outcome measures. Because prior studies have focused on variably defined outcome measures, the scope of these results has been limited, and it is difficult both to draw comparisons between cohorts and to apply the results to clinical practice. Given that children with polyarticular JIA continue to have active disease for the majority of their disease course, the development of continuous measures of disease activity is needed for a better description of the response to therapy in clinical trials, for defining and describing disease activity states at a single point in time and over time, for the translation of clinical trial results into the care of individual patients, and for the followup of individual patients in routine clinical care.

Currently, the primary outcome measure used in therapeutic trials in polyarticular JIA is the American College of Rheumatology (ACR) Pediatric 30 (Pedi 30) criteria for improvement (8). Developed in 1997, the ACR Pedi 30 was designed to distinguish between active treatment and placebo and is the only prospectively validated measure of disease activity in JIA. Although not prospectively evaluated, the ACR Pedi 20, Pedi 50, Pedi 70, and Pedi 90 measures are now also used as outcome measures in clinical trials. Preliminary definitions of disease flare and inactive disease in polyarticular JIA have also been proposed (9, 10). Furthermore, although the development of the ACR pediatric response measures was a significant advance for pediatric rheumatology, the utility of these measures is limited, because they assess the relative response (i.e., the change in disease status relative to a baseline clinic visit or other prior clinic visit), are dichotomous, and do not provide an absolute measure of the disease state.

Continuous measures of disease activity have been in use in adult rheumatoid arthritis (RA) since development of the Disease Activity Score (DAS) in 1993 (11). Subsequently, additional continuous measures have been proposed and validated in adult RA, including the DAS in 28 joints (DAS28) (12), the Simplified Disease Activity Index (SDAI) (13), and the Clinical Disease Activity Index (CDAI) (14). Although these measures have apparent face validity for polyarticular-course JIA, the utility of these measures in polyarticular-course JIA has not been extensively evaluated. We therefore sought to determine whether the DAS, DAS28, SDAI, and CDAI are potentially useful as continuous measures of disease activity in patients with polyarticular-course JIA, by measuring their predictive and discriminative abilities for the ACR pediatric measures of relative response. In this secondary analysis of 2 previously conducted clinical trials, disease activity was determined before and after treatment with etanercept in patients with polyarticular-course JIA.

PATIENTS AND METHODS

Patients.

We conducted a secondary analysis of data from 2 trials of etanercept in children and adolescents with polyarticular-course JIA (known as juvenile RA at the time when the trials were conducted). As a result of these trials, etanercept, a soluble tumor necrosis factor (TNF) receptor–Fc fusion protein, was approved by the US Food and Drug Administration for the treatment of moderately to severely active polyarticular-course JIA in patients ages 2 years and older (15). Approval for the analyses reported here was obtained from the Seattle Children's Hospital institutional review board.

Key eligibility criteria that were common to both studies included the following: a diagnosis of JIA according to the ACR (formerly, the American Rheumatism Association) criteria (16); a systemic, polyarticular, or pauciarticular disease onset, with a polyarticular disease course; ≥5 swollen joints accompanied by pain and/or tenderness and/or warmth, and ≥3 joints with limitation of motion at screening; a disease duration that was long enough for the patient to have received an adequate trial of nonsteroidal antiinflammatory drugs; normal hepatic, renal, and immunologic function; and no prior anti-TNF antibody therapy. Patients enrolled in the first study (17) were ages 4–17 years, could not have received methotrexate (MTX) for ≥14 days before receiving etanercept, could not have functional class IV disease according to the ACR criteria (18), and could not have tested positive for anti–double-stranded DNA antibodies. Patients in the second study (19) were ages 2–18 years, were receiving MTX at a stable dosage of 0.3–1.0 mg/kg/week at the time of randomization, and could not have received intraarticular glucocorticoid injections within 28 days prior to enrollment.

The first study was a phase II/III clinical trial conducted in 2 phases: an open-label phase and a double-blind, controlled phase. During the open-label phase, all patients (n = 69) received etanercept 0.4 mg/kg (maximum 25 mg) subcutaneously twice weekly for up to 3 months. After 3 months, patients who had achieved a clinical response were randomly assigned to receive etanercept 0.4 mg/kg or placebo (n = 51). Patients continued on this regimen for up to 4 months or until the occurrence of a disease flare. Patients who had some response but did not meet criteria to be considered a responder in the first open-label phase, patients with a disease flare during the double-blind phase, and patients who completed the double-blind phase were then eligible to participate in an open-label extension study for up to 120 months (n = 58).

In the second trial (phase III), patients (n = 25) were randomly assigned to receive MTX (at the dose received prior to study entry) once weekly plus either etanercept (0.4 mg/kg up to 25 mg) or placebo subcutaneously twice weekly for 6 months. After the controlled portion of the trial, all patients (n = 20) received etanercept plus MTX on an open-label basis for an additional 6 months. Patients were allowed to enter the open-label phase after 2 months of blinded treatment for disease flare or lack of response. The trial was stopped before meeting its enrollment goals because of challenges in patient recruitment.

In both trials, the response to treatment was assessed using the following pediatric core set components: physician's global assessment of disease severity (PhGA) as measured in centimeters on a 10-point visual analog scale (VAS), patient's/parent's global assessment of overall well-being (PtGA) as measured in millimeters on a 100-point VAS or in centimeters on a 10-point VAS, number of active joints (tender joint count [TJC] and swollen joint count [SJC]), number of joints with limitation of motion, Childhood Health Assessment Questionnaire (C-HAQ), and measurements of the C-reactive protein (CRP) level (mg/liter or mg/dl) or the erythrocyte sedimentation rate (ESR; mm/hour) (8).

Statistical analysis.

Pooled data from these trials were used to calculate values for the DAS and the DAS28, the SDAI, and the CDAI, using the following equations:

equation image
equation image
equation image
equation image

The DAS was calculated using a 53-TJC in place of the Ritchie Articular Index (20), based on the modified DAS formula described by Fransen et al (21). Because the SDAI is based on the CRP value, the versions of the DAS and DAS28 also based on the CRP value were chosen so that comparisons would be performed across scores based on the same marker of inflammation. Cutoffs for disease states were determined based on previously published cutoff points in adult RA (22, 23). The cutoffs for high disease activity were as follows: DAS >3.7, DAS28 >5.1, CDAI score >22, and SDAI score >26. Cutoffs for moderate disease activity were as follows: DAS >2.4 and ≤3.7, DAS28 >3.2 and ≤5.1, CDAI score >10 and ≤22, and SDAI score >11 and ≤26. Cutoffs for low disease activity were as follows: DAS ≤2.4, DAS28 ≤3.2, CDAI score ≤10, and SDAI score ≤11. Disease was considered to be in remission if the patient had a DAS <1.6, a DAS28 <2.6, a CDAI score ≤2.8, or an SDAI score ≤3.3 (24).

All children were analyzed together, irrespective of the treatment arm. Pearson's correlations were used to correlate the DAS, DAS28, CDAI, and SDAI values with the pediatric core set components. The ACR Pedi 30 was used as the standard to evaluate the agreement of disease improvement between the DAS and the pediatric outcome measures. Kappa statistics (25) were calculated between the ACR Pedi 30 and the European League Against Rheumatism (EULAR) response criteria based on the DAS and the DAS28 (26). Participants were dichotomized into levels of response using the EULAR response criteria. Children who achieved good or moderate levels of response were categorized as responders, and children who did not meet these criteria were categorized as nonresponders.

The areas under the curve (AUCs) of the receiver operating curve (ROC) (26) were calculated to assess the diagnostic accuracy of the adult RA measures for discriminating the ACR Pedi 30 response. ROCs are obtained by plotting the sensitivity of a test versus (1 − specificity) and are used to assess the diagnostic accuracy of a test. The AUC of an ROC reflects the discriminative properties of a test. AUC values range from 1.0 (perfect test) to 0.5 (useless test). In general, higher AUC values indicate better diagnostic properties (27).

RESULTS

Patients.

Ninety-four patients with JIA were included in these analyses. The baseline characteristics of the patients are shown in Table 1.

Table 1. Characteristics of the 94 patients at baseline*
  • *

    The analysis included all patients with at least 1 nonmissing baseline value or efficacy end point. JIA = juvenile idiopathic arthritis.

  • Data were available for 83 patients.

Age, mean ± SD years10.4 ± 4.1
Female sex, no. (%)56 (60)
Rheumatoid factor positive, no. (%)17 (21)
JIA onset type, no. (%) 
 Pauciarticular12 (13)
 Polyarticular56 (60)
 Systemic26 (28)

Disease activity score values and disease activity states.

At baseline, the median number of joints with active disease per patient was 27, and the mean C-HAQ score was 1.4. Patients entered in the trials had a history of 5.4 ± 3.8 years (mean ± SD) of disease and thus were considered to have established disease. The mean DAS, DAS28, CDAI score, and SDAI score at baseline were 3.7, 4.7, 30.8, and 36.4, respectively, and corresponded to a high disease activity state based on the CDAI and the SDAI, and moderate disease activity according to the DAS and the DAS28 (Table 2).

Table 2. Estimated responses based on the pediatric core set components and adult rheumatoid arthritis measures*
 Patients with JIA
Baseline (n = 94)Month 3 (n = 81)Month 6 (n = 39)
  • *

    Except where indicated otherwise, values are the number of patients/number of patients tested (%). All patients with ≥1 nonmissing baseline value or efficacy end point value were included. JIA = juvenile idiopathic arthritis; DAS = Disease Activity Score; DAS28 = DAS in 28 joints; CDAI = Clinical Disease Activity Index; SDAI = Simplified Disease Activity Index; IQR = interquartile range; C-HAQ = Childhood Health Assessment Questionnaire; CRP = C-reactive protein; ESR = erythrocyte sedimentation rate; ACR = American College of Rheumatology; EULAR = European League Against Rheumatism.

DAS, mean ± SD3.7 ± 1.42.2 ± 1.31.7 ± 1.4
DAS28, mean ± SD4.7 ± 1.33.1 ± 1.32.6 ± 1.3
CDAI, mean ± SD30.8 ± 12.514.9 ± 11.19.9 ± 12.5
SDAI, mean ± SD36.4 ± 16.917.7 ± 14.211.1 ± 14.0
Total number of joints with active disease, median (IQR) (range 0–71)27 (16–35)10 (4–22)3 (1–11)
Joints with loss of movement, median (IQR) (range 0–71)22 (12–32)15 (8–24)6 (4–25)
Physician's global assessment, median (IQR) (range 1–10)6.0 (5.0–8.0)2.0 (1.0–4.0)1.0 (1.0–2.0)
Patient's global assessment, median (IQR) (range 1–10)5.0 (3.0–7.0)2.0 (1.0–4.0)1.0 (1.0–3.0)
C-HAQ score, mean ± SD1.4 ± 0.90.9 ± 0.80.6 ± 0.8
CRP, mean ± SD mg/dl6.0 ± 7.93.0 ± 5.81.4 ± 2.5
ESR, mean ± SD mm/hour41.0 ± 29.624.9 ± 26.414.2 ± 15.3
ACR Pediatric 2070/81 (86.4)33/39 (84.6)
ACR Pediatric 3064/81 (79.0)33/39 (84.6)
ACR Pediatric 5049/81 (60.5)32/39 (82.1)
ACR Pediatric 7035/81 (43.2)23/39 (59.0)
ACR Pediatric 909/81 (11.1)9/39 (23.1)
EULAR responder based on DAS59/79 (74.7)34/35 (97.1)
EULAR responder based on DAS2866/79 (83.5)32/35 (91.4)

At the 3-month followup, the mean number of joints with active disease per patient decreased from 27 to 10. The mean C-HAQ score decreased from 1.4 to 0.9, for a mean change in the C-HAQ score of −0.5, indicating a greater than minimum clinically important difference for improvement from baseline, which has been shown to be a change in the C-HAQ score of −0.188 (28). The mean values for the DAS/DAS28 also decreased and corresponded to a low disease activity state (with improvement of >1.2 in the DAS28 indicating a clinically meaningful response to therapy) (22). The mean values of the CDAI/SDAI corresponded to a state of moderate disease activity. At the 6-month followup, the mean number of joints with active disease per patient was 3, and the mean C-HAQ score was 0.6. The mean DAS/DAS28/CDAI score for this visit corresponded to low or minimally detectable disease activity, and the mean SDAI score corresponded to moderate disease.

At 3 months, a total of 69 patients had been treated with etanercept alone, 13 patients had been treated with etanercept plus MTX, and 12 patients received MTX alone. At 6 months, 25 patients were receiving etanercept alone, 26 patients were receiving no biologic or traditional disease-modifying antirheumatic drug, 13 patients were receiving etanercept plus MTX, and 12 patients were receiving MTX alone. The majority of patients achieved a response level of good (decrease in the DAS or DAS28 of ≥1.2, and DAS ≤2.4 or DAS28 ≤3.2) or moderate (decrease in the DAS or DAS28 of >0.6 but ≤1.2 and DAS ≤3.7 or DAS28 ≤5.1, or decrease in the DAS or DAS28 of >1.2 and DAS >2.4 and ≤3.7 or DAS28 >3.2 and ≤5.1) by 3 and 6 months, as defined by the EULAR criteria based on the DAS or DAS28 (22, 23).

The sensitivity of the EULAR response (good or moderate) for the ACR Pedi 30 was high at both 3 months and 6 months, indicating that if a EULAR response was achieved, then an ACR Pedi 30 response was also likely, although the specificity was substantially lower at both time points, indicating that not achieving a EULAR response was not always associated with nonresponse for the ACR Pedi 30 (Table 3). The positive predictive value for a EULAR response based on the DAS was 0.93 at 3 months and 0.91 at 6 months, whereas the negative predictive value was only 0.55 at 3 months but was 1.0 at 6 months.

Table 3. Agreement between the EULAR response based on the DAS, DAS28, and ACR Pedi 30 criteria*
EULAR responseNo. of patientsκ (95% CI)SensitivitySpecificityPositive predictive valueNegative predictive valueAccuracy
  • *

    The number of patients includes those with nonmissing values for both the Disease Activity Score (DAS) or DAS in 28 joints (DAS28) and the American College of Rheumatology (ACR) Pediatric 30 (Pedi 30). Accuracy was calculated as the proportion of patients with a correct diagnosis (“yes” for both the ACR Pedi 30 and European League Against Rheumatism [EULAR] response or “no” for both the ACR Pedi 30 and EULAR response). 95% CI = 95% confidence interval.

Based on DAS and ACR Pedi 30       
 Month 3790.5 (0.3, 0.7)0.860.730.930.550.84
 Month 6350.4 (−0.2, 0.9)1.000.250.911.000.91
Based on DAS28 and ACR Pedi 30       
 Month 3790.5 (0.2, 0.7)0.920.530.890.620.85
 Month 6350.5 (0.1, 1.0)0.970.500.940.670.91

Correlations between the DAS and the pediatric core set components.

The correlation between the continuous outcome measures and each of the pediatric core set components was moderate to very good (Table 4). As anticipated, because the swollen and tender joint counts make the greatest contribution to each DAS equation, the closest correlation was measured for the total count of joints with active disease (r = 0.927 at 6 months) (Figure 1).

Table 4. Correlations between disease activity scores/indices and other indicators of disease activity*
 No. of patientsC-HAQPhysician's global assessmentPatient's global assessmentNo. of joints with active diseaseCRP
  • *

    Values are the correlation coefficients (95% confidence intervals). The Childhood Health Assessment Questionnaire (C-HAQ) scores 8 domains on a 4-point scale (0–3), with higher numbers representing greater disability. The physician's and patient's global assessments were based on a scale of 0 (asymptomatic) to 10 (severe). The C-reactive protein (CRP) level is measured in mg/dl. DAS = Disease Activity Score; DAS28 = DAS in 28 joints; CDAI = Clinical Disease Activity Index; SDAI = Simplified Disease Activity Index.

  • Number of patients with nonmissing disease activity score/index.

DAS      
 Baseline930.53 (0.36, 0.66)0.51 (0.35, 0.65)0.58 (0.42, 0.70)0.74 (0.63, 0.82)0.52 (0.35, 0.65)
 Month 3800.58 (0.42, 0.71)0.64 (0.48, 0.75)0.65 (0.50, 0.76)0.79 (0.69, 0.86)0.47 (0.28, 0.63)
 Month 6360.62 (0.37, 0.79)0.68 (0.45, 0.83)0.65 (0.41, 0.81)0.93 (0.86, 0.96)0.57 (0.30, 0.76)
DAS28      
 Baseline930.60 (0.45, 0.71)0.61 (0.46, 0.72)0.68 (0.55, 0.77)0.62 (0.47, 0.73)0.68 (0.55, 0.78)
 Month 3800.62 (0.46, 0.74)0.68 (0.53, 0.78)0.67 (0.53, 0.78)0.72 (0.59, 0.81)0.62 (0.46, 0.74)
 Month 6360.66 (0.42, 0.81)0.67 (0.44, 0.82)0.59 (0.33, 0.77)0.91 (0.82, 0.95)0.66 (0.43, 0.81)
CDAI      
 Baseline940.49 (0.32, 0.63)0.55 (0.39, 0.68)0.55 (0.39, 0.68)0.85 (0.79, 0.90)0.43 (0.25, 0.58)
 Month 3810.56 (0.39, 0.69)0.73 (0.60, 0.81)0.71 (0.58, 0.80)0.83 (0.75, 0.89)0.37 (0.17, 0.55)
 Month 6390.60 (0.34, 0.77)0.77 (0.60, 0.87)0.58 (0.32, 0.76)0.97 (0.93, 0.98)0.54 (0.26, 0.74)
SDAI      
 Baseline930.58 (0.43, 0.70)0.62 (0.47, 0.73)0.60 (0.45, 0.71)0.74 (0.64, 0.82)0.77 (0.67, 0.84)
 Month 3800.63 (0.48, 0.75)0.72 (0.60, 0.81)0.63 (0.48, 0.75)0.76 (0.65, 0.84)0.70 (0.56, 0.80)
 Month 6360.70 (0.48, 0.84)0.71 (0.50, 0.84)0.58 (0.31, 0.76)0.94 (0.89, 0.97)0.66 (0.43, 0.81)
Figure 1.

Scatter plots showing the correlations between the number of joints with active disease per patient and the Disease Activity Score (DAS) at baseline (A) and after 3 months (B) and 6 months (C) of etanercept therapy.

AUCs of ROCs for the disease activity scores and the pediatric ACR measures of response.

The AUCs of the ROC for each disease activity score and each ACR Pedi 30 response were >0.75, indicating good convergent validity between the disease activity scores and the ACR Pedi 30 criteria (Table 5). There was a trend toward the AUCs of the ROC being higher for the ACR Pedi 30, Pedi 50, and Pedi 70 than for the ACR Pedi 20 and ACR Pedi 90.

Table 5. Concordance between disease activity scores and the ACR Pediatric 30 criteria*
 Month 3Month 6
  • *

    Values are the areas under the curve (95% confidence intervals) of the receiver operating curve. ACR = American College of Rheumatology; DAS = Disease Activity Score; DAS28 = DAS in 28 joints; CDAI = Clinical Disease Activity Index; SDAI = Simplified Disease Activity Index.

DAS0.82 (0.69, 0.94)0.85 (0.64, 1.05)
 Improvement from baseline0.85 (0.74, 0.96)0.79 (0.54, 1.04)
DAS280.82 (0.69, 0.94)0.89 (0.76, 1.02)
 Improvement from baseline0.87 (0.77, 0.96)0.95 (0.87, 1.03)
CDAI0.77 (0.64, 0.91)0.89 (0.74, 1.04)
 Improvement from baseline0.88 (0.79, 0.97)0.98 (0.96, 1.01)
SDAI0.76 (0.63, 0.90)0.86 (0.67, 1.04)
 Improvement from baseline0.90 (0.82, 0.99)0.93 (0.78, 1.08)

DISCUSSION

In this sample of children with polyarticular-course JIA enrolled in clinical trials of etanercept, the DAS, DAS28, SDAI score, and CDAI score correlated well with the individual components of the pediatric core set of criteria. There was an overall decrease in the mean level of disease activity, as defined by the DAS, DAS28, SDAI score, and CDAI score, between baseline and 6 months, and the majority of children achieved a good or moderate EULAR response by 3 months of followup. Furthermore, the AUCs of the ROC for each of the continuous measures of disease activity and each ACR pediatric measure of relative response were high.

These data support the prior study by Lurati and colleagues that retrospectively examined the concordance between the ACR Pedi 30 and ACR Pedi 20, and the DAS, DAS28, and EULAR categories of response in a cohort of 75 children with multiple JIA types treated with MTX or a TNF inhibitor (29). In this cohort, the highest level of concordance (71%) was measured between the DAS and the ACR Pedi 30 (AUC = 0.735). This relationship did not appear to be significantly different for children who were younger than age 16 years compared with those who were older than age 16 years.

Together, these data indicate that these continuous measures of disease activity are potentially applicable to polyarticular-course JIA. In contrast to the ACR pediatric measures of relative response, the continuous measures have the advantage of describing not only change in disease activity, but also the disease state at a single point in time. These measures are therefore useful in both the research and clinical settings, and the availability of a continuous measure of disease activity in JIA would be a valuable tool, facilitating comparison of patient disease status and treatment response across clinical trials and their translation to routine clinical care.

The process of validation of a new outcome measure requires formal testing of the different components of validity: face validity, feasibility, content validity, criterion validity, discriminant validity, and construct validity (30). Our analyses and those of Lurati and colleagues (29) suggest good discriminant validity for these scores in polyarticular-course JIA, as measured by the AUCs of the ROCs for each score relative to each ACR pediatric measure of relative response. In addition, these current analyses indicate good construct validity for the DAS, DAS28, SDAI, and CDAI in polyarticular-course JIA, with good correlation between the DAS, DAS28, SDAI, and CDAI and the components of the pediatric core set. The DAS, DAS28, SDAI, and CDAI also have apparent face validity for polyarticular-course JIA, because they capture the majority of content measured by the pediatric core set components. Nevertheless, additional work will be required to determine whether the exclusion of joints with loss of range of motion, exclusive of measures of disability, and the use of reduced joint counts (i.e., the 28- and 44-joint counts), significantly affect a provider's perception of the face validity of a measure. Most pediatric rheumatologists would not equate DAS or DAS28 remission with clinical remission, because of the number of joints with active disease that persist in patients who meet the criteria for DAS or DAS28 remission. In our analyses, SDAI and CDAI remission appeared to be more stringent (median 0–1 active joints).

We were not able to assess all aspects of the validity of these scores in this cohort, and the face validity, feasibility, and criterion validity of these scores will require additional study. These analyses were also limited by the sample size, particularly at the 6-month time point, which likely affected the precision of our estimates and may have also resulted in higher correlation and AUC measurements than would be seen in cohorts with larger numbers of children and more variable responses to treatment. In addition, the children enrolled in these protocols had severe disease at baseline, as evidenced by their total counts of joints with active disease, and these scores may not provide similar results in cohorts of children with less severe disease for whom changes in disease activity may not be as large.

The lower correlation between these scores and the 2 patient- and parent/proxy- reported outcomes (the C-HAQ and patient's global assessment of disease activity) requires further exploration as well, particularly given the known confounding effects of non–disease activity–related factors on the C-HAQ (31). Correlations were lowest between the disease activity measures and the C-HAQ, likely because there is no discrete measure of disability in the adult RA measures, and because the C-HAQ also reflects disability due to damage from prior disease activity as well as ongoing, active disease.

In contrast with the consistently high positive predictive value (proportion of patients with positive test results and a correct diagnosis), the negative predictive value (proportion of patients with negative test results and a correct diagnosis) for a EULAR response based on the DAS was low at the 3-month time point and perfect at the 6-month time point. The most likely explanation for this observation is the small sample size, especially at 6 months.

Despite these limitations, these data indicate that the DAS, DAS28, SDAI, and CDAI may be valid measures of disease activity in children with polyarticular-course JIA and support the additional validation of these scores in larger cohorts of children with polyarticular-course JIA that include children with early JIA (disease for <1 year) and with varying degrees of disease severity.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Ringold had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ringold, Chon, Singer.

Acquisition of data. Ringold, Singer.

Analysis and interpretation of data. Ringold, Chon, Singer.

ROLE OF THE STUDY SPONSOR

The study sponsors were not involved in the design or interpretation of the data. Publication of this article was not contingent on approval by the study sponsors.

Acknowledgements

We thank the Pediatric Rheumatology Collaborative Study Group investigators who participated in the original trials and Norman Ilowite, MD, for his contributions to the execution of one of the original studies included in our data analysis. We also thank Nan Zhang, PhD, of Amgen Inc. for support with the statistical analysis, Michele Hooper, MD, of Amgen Inc. for helpful discussions, and Julia R. Gage, PhD, whose work was funded by Amgen Inc., for editorial assistance.

Ancillary